RE: Suggestion to change RDF encoding of HasKey axioms [RE: I've implemented the change to the syntax of HasKey]

Short story: I am withdrawing my suggestion. Read below!

>-----Original Message-----
>From: Ivan Herman [mailto:ivan@w3.org]
>Sent: Saturday, March 21, 2009 3:21 PM
>To: Michael Schneider
>Cc: W3C OWL Working Group; Boris Motik
>Subject: Re: Suggestion to change RDF encoding of HasKey axioms [RE:
>I've implemented the change to the syntax of HasKey]
>
>Ouch. I fell into the same trap again:-( Yes, I remember the QCR
>problem!
>
>I must admit, having thought through all this: I do not like it.
>Sorry....:-( What this solution means is that if I have only object
>properties p1 an p2 as keys (or properties that I do not care the type
>of) I still have to say
>
>[ a owl:Key ;
>  owl:keyClass <C> ;
>  owl:keyObjectProperties (p1 p2) ;
>  owl:keyDataProperties ()
>]
>
>Ie, I _have_ to put the empty list for data properties even if, at that
>point, I do not really care about data properties. This is highly error
>prone, let alone ugly in RDF. Error prone because we can bet that people
>will forget to add that empty list, in which case the keys would not
>work.

I agree that this is error prone. In this case, the semantic condition for keys would /not/ apply, and the presumed key axiom would not be more than an arbitrary set of triples. I am now inclined to *not* make the suggested change.

>What I could imagine is:
>
>1. keep the old C owl:hasKey (p1 p2 p3 ... ) encoding with the
>assumption that the backward mapping to FS maps p1, p2, ... p3 to object
>properties
>2. _add_ the encoding you propose as an alternative in case I really
>really care that about the category of the properties, ie, I want to
>have an ontology that would be DL.

Puh, is this ugly!

And, what should the forward mapping do in such a case? Should it produce encoding #1 if the second set is empty on the FS side, and encoding #2 otherwise?

>But I do not think I would like to have alternative #1 only...
>
>Why did we have this whole thing again? Oh yes, the comment of Matthew
>on explicit typing of properties. Sigh... Maybe we should re-discuss
>this? (/me ducks:-)

We could say that we are consequent and perform no changes to the RDF encoding (just as for all other cases, when the FS became "fully typed"). As I said, I won't press any changes, and I can pretty well live with the status quo concerning the RDF-Based Semantics (at least, I could live with it in the past, and the situation has not changed for OWL 2 Full since then).

What I did not considered so far: Future working groups will, of course, have the chance to largely revise the structural specification, and then really allow for ideas like that of "inlined entity declarations". So, in principle at least, no doors really close for object/data property punning in a future version of OWL, even with the current RDF encoding.

So I hereby withdraw my suggestion! This does, however, not mean that other people may still go on pursuing this topic.
 
Michael

--
Dipl.-Inform. Michael Schneider
Research Scientist, Dept. Information Process Engineering (IPE)
Tel  : +49-721-9654-726
Fax  : +49-721-9654-727
Email: michael.schneider@fzi.de
WWW  : http://www.fzi.de/michael.schneider
=======================================================================
FZI Forschungszentrum Informatik an der Universität Karlsruhe
Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
Tel.: +49-721-9654-0, Fax: +49-721-9654-959
Stiftung des bürgerlichen Rechts, Az 14-0563.1, RP Karlsruhe
Vorstand: Prof. Dr.-Ing. Rüdiger Dillmann, Dipl. Wi.-Ing. Michael Flor,
Prof. Dr. Dr. h.c. Wolffried Stucky, Prof. Dr. Rudi Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
=======================================================================